US government disables Anthropic's Fable 5 and Mythos 5 worldwide
Anthropic had to pull the plug on its advanced AI systems, Fable 5 and Mythos 5, after the US government raised national security concerns.
Officials feared that these models' safety features could be bypassed, which might pose serious risks.
Negotiations didn't work out, so the government used export controls to disable them worldwide.
White House gave Anthropic 90 minutes
Amazon CEO Andy Jassy flagged vulnerabilities in the models at the White House's request, and the NSA backed up those worries after a review.
Anthropic's CEO Dario Amodei defended his tech, saying the risks were misunderstood.
Anthropic said no universal jailbreaks had been found. Still, with unresolved issues, the White House gave Anthropic just 90 minutes to take its systems offline, showing how seriously officials are taking AI safety right now.